JP2534446Y2 - Apparatus for determining valid samples from digitized signals. - Google Patents

Abstract

Improved significant sample detection for a pitch detector for use with speech analysis and synthesis methods by performing a reverse order search and a forward order search of digitized speech samples. A reverse search detector (12) is responsive to segmented digital samples for determining a set of candidate samples by initially selecting one of the digitized samples as a present candidate sample and comparing in reverse order each of the digitized samples with the present candidate sample until a digitized sample is found whose amplitude is greater than the present candidate sample or the compared sample is greater than a predefined number of samples from the present candidate sample. When either of the previous conditions occurs, the compared digital sample becomes the new present candidate sample and the reverse search continues. During the reverse search, each of the compared samples that has not replaced the present candidate sample is set equal to zero. After the reverse search has been performed and a set of candidate samples has been determined, a forward search detector (14) then initially determines a present significant sample. The latter detector compares this significant sample with each of the candidate samples until a candidate sample is found whose amplitude is greater than the present significant sample or the compared candidate sample is more than a predefined number of samples away from the present significant sample. When either of those conditions occurs, the forward search detector saves the value of the amplitude and location of the candidate sample and replaces the present significant sample with that candidate sample and continues the search. A single forward and reverse search determines all of the significant samples.

Description

[Detailed description of the invention]

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention generally relates to digitally encoding a human voice signal, compactly storing or transmitting the signal, and further synthesizing the signal. Related to doing.

[0002]

SUMMARY OF THE INVENTION The number of bits per second required to store or transmit coded human speech is less than the number of bits required to store or transmit voice using conventional pulse code modulation. Known human speech coding techniques are known. In order to use such a coding method that minimizes the number of bits, an analog audio sample is usually divided into time frames or segments each having a time width of about 20 ms before final encoding. is there. Normally, the sampling of the audio signal is performed at a rate of 8 kilohertz (KHz), and each sample is encoded into a multi-bit digital number. Further, the successively encoded samples are processed by a linear predictive coder (LPC), which determines appropriate filter parameters that simulate the formant structure of the vocal tract transfer function. Using these filter parameters, the current value of each signal sample can be effectively evaluated based on the weighting of a preselected number of previous sample values.

[0003] As a result of analysis, the voice signal is considered to be composed of an excitation signal and a formant transfer function. The excited component is a larynx or voice
box), and the formant transfer function results from the action on the excited components of the rest of the vocal tract. The latter component is further classified as voiced or unvoiced according to whether there is a fundamental frequency transmitted to the airflow by the vocal cords. If the excitation is unvoiced, the excitation component simply gives white noise. On the other hand, when there is a fundamental frequency transmitted to the airflow by the vocal cords, this excitation component is classified as voiced. The pitch detection, that is, the problem of determining the fundamental frequency of the voiced excitation component, which is a key parameter, is difficult to implement with a minimum amount of calculation.

US Pat. No. 4,561,102 shows one way to determine this pitch. The method used in this patent places a set of valid samples within a speech frame, first scanning all samples until the largest sample is found, and then scanning until the second largest sample is found. It is configured to iterate the search for samples. This process continues until a predetermined number of samples are found in the audio frame. This approach requires that the number of scans that must be performed is proportional to the square of the number of samples to be found.

[0005] However, this method has the disadvantage that it takes a long time to find extremely many samples. This method uses a digital signal processor (DS) for some uncomplicated coding schemes.
P), but when the encoding method becomes more complicated,
In this type of signal processor, there is a problem that if the above-described specific search method is applied to each frame, the calculation capability is insufficient.

An object of the present invention is to solve the disadvantages of the prior art.

[0007]

SUMMARY OF THE INVENTION To this end, an apparatus according to the present invention is configured to utilize an inverse order search detector and a forward order search detector to determine valid samples in a speech signal in response to the signal. .

The inverse search detector determines a set of candidate samples in response to a segment of the digitized speech signal, which first selects one of the digitized samples as a current candidate sample, Next, by comparing each digitized sample in reverse order with the current candidate sample until finding a digitized sample whose amplitude is greater than that of the current candidate sample,
Alternatively, the comparison is performed until the compared sample is larger than a predetermined number of samples from the current candidate sample. If any of the previous conditions occur, the compared sample becomes the new current candidate sample, and the reverse search continues. During this reverse search,
Each of the comparison samples that are not current candidate samples is set equal to zero.

[0009] Once the reverse order search has been performed and a set of candidate samples has been determined, then the forward search detector first determines the current valid sample from these candidate samples. This latter detector compares the current valid sample to each of the candidate samples until a candidate sample whose amplitude is greater than the current valid sample is found, or the comparison is made by comparing the compared candidate sample to the current valid sample. Until the number of samples exceeds a predetermined number of samples. When any of these conditions occur, the forward search detector shifts the amplitude and position values of the candidate sample, and replaces the current valid sample with the candidate sample,
Continue searching.

[0010]

FIG. 1 illustrates a maximum locator which is a feature of the present invention. The maximum locator determines valid samples in response to a frame of digital samples representative of the analog audio signal received via path 11. These audio frames are preprocessed as follows. To reduce noise, the speech is first reduced-leaked, ie, low-pass filtered, and then digitized and quantized. This digitized speech is then divided into 20 ms frames, each frame
For example, it is composed of 160 samples. Further, as will be apparent to those skilled in the art, the maximum locator can respond to other forms of signal derived from the analog audio signal that can be used to determine pitch. One such signal is a normal order prediction error or a residual signal obtained when calculating the LPC coefficient.

Here, the operation of the maximum locator 10 of FIG. 1 will be described in detail. This locator provides an output signal on path 17 shown in FIG. 4 in response to the audio frame samples shown in FIG. The reverse order search detector 12 detects the sample shown in FIG. Only a subset of the 160 samples is shown in the figure. Detector 12 starts at sample 159 and searches from right to left, performing the following operations.
Detector 12 assumes sample 159 as the current candidate sample and stores its value. The detector 12 then examines each sample to the left until it encounters another sample with an amplitude greater than the current candidate sample or another sample which is the nineteenth sample from the current candidate sample being examined. . If you encounter a sample with large amplitude,
Alternatively, when the number of examined samples equals 19 from the current candidate sample, detector 12 stores the sample as a new candidate sample and repeats the previous search procedure. The basis for terminating the search after obtaining 19 samples and starting a new search is based on the assumption that the highest pitch obtained with human speech is about 420 Hz, which yields 19 samples at a sample rate of 8 KHz. As the detector 12 examines each sample, if that sample is less than or equal to the current candidate sample and one of the current candidate samples
If so, the sample under investigation is set to zero.

Here, a method in which the detector 12 processes the sample shown in FIG. 2 to generate the sample shown in FIG. 3 will be described. Detector 12 starts with sample 159,
Proceed to the left while examining each successive sample. For example, sample 158 is smaller than 159, where sample 158 is set equal to zero. When detector 12 encounters sample 152, detector 12 determines that the amplitude of this sample is greater than that of sample 159. The detector then re-initializes the search procedure using sample 152 as the current candidate sample. The search then proceeds from sample 152 until it reaches sample 133. Sample 133 is from samples 152 to 19
Since it is the sample destination, sample 133 is used as the current candidate sample, and the search proceeds to the left. FIG. 3 shows the result of the detector 12 that advances the search to the left and zeros out samples that do not satisfy the above search procedure.

The forward search detector 14 detects the output of the reverse search detector 12 and performs the following search procedure from left to right. Detector 14 starts at sample 0 and uses sample 0 as the current valid sample, until a sample larger than the current valid sample is obtained, or until a sample greater than 18 in number from the current valid sample is examined. Each of the samples received from the inverse search detector 12 is searched. This is set to zero if the sample examined does not satisfy one of the criteria already mentioned. When a sample satisfies this criterion, the amplitude and position of the sample are stored, and this sample becomes the new current valid sample.

Here, the response of the detector 14 to the sample shown in FIG. 3 will be described. Detector 14 starts at sample 0 and searches until it exceeds sample 18, which is 18 samples. Sample 19 is recorded as the current valid sample. When detector 14 searches from sample 104, no sample larger than sample 104 is obtained, sample 123 is made the current valid sample, and the search proceeds from sample 123. FIG. 4 shows the result of the forward search detector 14. Note that some samples with zero values here are nonetheless shown as valid samples, but not shown in FIG. These zero samples are later rejected by the threshold detector 16.

That is, detector 16 detects the samples shown in FIG. 4 and rejects all samples that are not greater than 25 percent of the maximum sample amplitude. The threshold detector 16 first determines the maximum sample amplitude, and then rejects all samples whose amplitude does not exceed 25 percent of this maximum amplitude.

FIG. 5 shows a flowchart of a program used to control the digital signal processor to perform the functions of the detectors 12, 14, 16. FIG. 6 shows such a digital signal processor system. The system shown in FIG.
Performs tasks that require low pass filters and D / A conversion. Further, the system provides a known program for performing the segmentation of the digital samples received from converter 612 into frames. The digital signal processor 601 performs various functions described above using the PROM 602 and the RAM 603. The program stored in the PROM 602 executes the flowchart shown in FIG.

Here, the program shown in FIG. 5 will be described in detail. Blocks 501-507 implement the inverse search detector 12. Blocks 501 and 502 are used for setting up two indices j and i.
The constant L is set equal to the number of samples, which in this embodiment is 160 samples. Next, the program executes blocks 503-507 until all samples have been examined.
Proceed to go around. These samples are contained in the array denoted by γ. The decision block 504 determines whether the amplitude of the current sample under investigation is smaller than the amplitude of the current candidate sample and exceeds the area of 18 samples. If both of these conditions are met, block 503 sets the current sample under investigation to zero.
Is executed. If the current sample under investigation is greater than or equal to the current candidate sample, or exceeds the range of 18 samples, the current sample becomes the new current sample. Block 506 simply decrements the index used to cycle through all samples, and decision block 507 determines whether all samples have been examined.

Blocks 508-515 implement the forward search detector 14. This detector determines the valid sample and stores the amplitude of the sample in array a and the position of the sample in array d, both of which are indexed by n. Blocks 508, 509, 510 set up the initial value of the index. Decision block 5
11 determines whether the sample currently being investigated is greater than the current valid sample, or whether the range of samples from the current valid sample is greater than 18 samples. If any of these conditions are true, block 512 is executed, resulting in a new current valid sample made equal to the sample currently being examined, and placing this sample in arrays a and d. Finally, block 512 increments index n. If these conditions are not met, block 513 is executed to zero the sample under investigation. Block 514 increments index i. Decision block 515 determines whether all of the samples have been examined.

The embodiments described above are merely illustrative of the principles of the present invention, and it will be apparent to those skilled in the art that other configurations are possible without departing from the spirit and scope of the present invention.

[Brief description of the drawings]

FIG. 1 is a schematic diagram of a maximum locator according to the present invention.

FIG. 2 is a diagram of an input digitized audio signal.

FIG. 3 is a diagram of an audio signal after being processed by the inverse search detector of FIG. 1;

FIG. 4 is a diagram of the sample of FIG. 3 after being processed by the forward search detector of FIG. 3;

FIG. 5 is a flowchart of a program for executing the maximum locator of FIG. 1;

FIG. 6 is a schematic diagram of the digital signal processor of FIG. 1;

[Explanation of symbols]

 12 reverse order search detector 14 forward order search detector 16 threshold value detector 601 digital signal processor 612 A / D converter

Claims (4)

(57) [Scope of request for utility model registration] 1. A digitized speech signal that is a potential pitch pulse in response to a digitized speech signal comprising a plurality of input speech segments, each having a plurality of non-zero samples. A set of important samples from the set, the samples of the set being of the largest magnitude, and being at least equal to the highest pitch encountered in human speech, at least in a predetermined time interval, In a maximal locator that is away from the samples, searching backwards for one sample of the input speech segment and selecting the first sample encountered in the search and another sample satisfying the following criteria: A first searcher (12), wherein the sample size is the most recently selected sample If the distance is greater than the selected sample, the sample is selected. If the sample is separated from the recently selected sample by a predetermined interval, the first searcher is configured to select the sample. A second searcher (14) that searches forward in the searched samples and selects the first sample encountered in the search and other samples satisfying the following criteria: A second searcher adapted to select the sample if larger than the recently selected sample and to select the sample if it is at least a predetermined distance from the recently selected sample; A third searcher for searching the samples selected by the two searchers to select a set of important samples, the important samples being the largest size samples; A maximal locator comprising a sampler and a third searcher consisting of all samples having a size that is a predetermined percentage of the largest size. 2. The maximal locator of claim 1, wherein the first searcher sets a non-selected sample to zero when it encounters it. 3. The maximal locator of claim 2, wherein the second searcher sets a non-selected sample to zero when it encounters it. 4. The maximal locator according to claim 1, wherein said second searcher stores a size and a position of each selected sample.